// YOUR PRESSURE WITH ITS REAL MARGIN (±PSI) · BERTO 15% DROP MODEL · 321-TIRE BRR CENSUS · ETRTO CAPS
A nominal 28 mm road tire measures on average 28.4 mm on an 18 mm internal rim and ~29.5 mm on a 21 mm rim. We measured it over BicycleRollingResistance's census of 159 road tires: modern nominals already assume a wide rim (ETRTO 2020).
Because gravel manufacturers spec the nominal on 21–25 mm rims. In BRR's census of 162 gravel tires (17.8 mm rim), a nominal 40 mm measures on average 38 mm — 2 mm less than what's printed on the sidewall. No calculator on the market corrects this with data.
Real tire width: it explains between 51% and 68% of the output variance on the road (Sobol analysis, 81,920 evaluations). Rider weight, against intuition, explains less than 9%. That's why this calculator asks for width to 1 mm and is relaxed about weight.
At gravel pressures (30–40 PSI), a typical gauge (±5%) is the system's largest error source: up to 73% of total variance per our sensitivity analysis. At those pressures your gauge matters more than the model — use a low-range digital one.
72.5 PSI (5 bar) per ETRTO, no exceptions — and 60 PSI if the rim is ≥30 mm internal. Exceeding it risks sudden blow-off. BikeLab PSI Calculator applies these caps automatically and shows the calculated value next to the capped one.
Because we ran a global sensitivity analysis (Sobol indices, 81,920 evaluations per configuration) on the model's variables, and the result was unambiguous: on the road, real tire width explains between 51% and 68% of the output variance — more than weight, load distribution and everything else combined. Pressure depends on width squared (the L/W² term): a 1.5 mm error in width moves the result more than a 2 kg error in your weight.
That's why we're relaxed about weight (±2 kg barely matters: under 9% of the variance) but serious about width. And that's why the optional caliper-measured width field exists: if you measure your mounted, inflated tire, that input's uncertainty drops from ±1.5 mm to ±0.5 mm and your result's band shrinks to a third. On gravel the dominant factor is something else entirely: your pressure gauge (up to 73% of the variance at low pressures) — see the warning in the results.
No calculator on the market explains why it asks what it asks. We publish the analysis.
The tire should deflect 15% of its section height under load. Frank Berto (technical editor of Bicycling Magazine) measured 50 tires at 7 pressures (40–160 PSI) and 8 loads (20–220 lbs per wheel) on 700C wheels. Our core formula is an empirical fit to his chart, accurate to ±2 PSI within its domain:
Per-wheel load comes from total weight and geometry: racing 40/60 (front/rear), endurance 45/55, city/touring 35/65.
The rear is computed with pure Berto on its static load. The front is not: under braking, load transfers massively to the front wheel, and a soft front degrades handling. So the front is coupled to the rear: Pfront = 0.93 · Prear, with k calibrated by sweep against SILCA and Rene Herse across 6 validation cases. It is the same pattern both references apply without declaring it.
The width printed on the sidewall is not the real mounted width. We captured BicycleRollingResistance's complete nominal-vs-measured census — 159 road tires (18 mm internal rim) and 162 gravel tires (17.8 mm) — and fit by regression:
The 0.4 mm-per-rim-mm slope was verified with the only tire BRR measured on two different rims (+11% deviation, within criterion). Census finding: modern gravel tires measure on average 2 mm less than nominal on a narrow rim — manufacturers today spec on wide rims (ETRTO 2020), not on the 13 mm rims of the Berto era. The classic rule corrected twice what the manufacturer had already corrected; ours is anchored to 2014–2026 data.
If a cap limits your result, we show both values: calculated and applied, with an explanation.
The band (±PSI) is the 90% confidence interval propagated via Latin Hypercube Sampling (10,000 samples per configuration) over input uncertainty: weight ±2 kg, real width ±1.5 mm (±0.5 mm if caliper-measured), load distribution ±3 points. It does not include your gauge error (typically ±5%): that is your instrument's uncertainty, not the model's, and mixing them would make honest output look imprecise about something it doesn't control. On gravel, at low pressures, the gauge is the largest real error source — use a digital one.
We ran 6 cases (70–100 kg, 25–45 mm, road and gravel) through the SILCA and Rene Herse calculators live, and compared against the average of both. Result: 9 of 12 wheels within ±10% (road criterion) or gravel's widened tolerance (the two references diverge from each other by up to 30% on gravel). The 3 remaining deviations have identified, documented causes (references rounding to integer widths, and the weight scaling described below).
| Case | Config | BLS F/R | SILCA F/R | R.Herse soft–firm |
|---|---|---|---|---|
| 1 | 83 kg · 28 mm · road · tubeless · 21 rim | 68 / 73 | 70 / 71.5 | 54–67 |
| 2 | 95 kg · 28 mm · road · tubeless · 21 rim | 78 / 84 | 71.5 / 73.5 | 61–76 |
| 3 | 75 kg · 32 mm · road · tube · 19 rim | 58 / 62 | 63.5 / 65 | 46–57 |
| 4 | 85 kg · 40 mm · gravel · tubeless · 24 rim | 35 / 38 | 34.5 / 36 | 34–42 |
| 5 | 100 kg · 45 mm · gravel · tubeless · 25 rim | 36 / 39 | 29.5 / 30.5 | 35–43 |
| 6 | 70 kg · 25 mm · road · tubeless · 21 rim | 65 / 70 | 80.5 / 83 | 55–72 |
For riders over 90 kg this calculator recommends firmer pressures than SILCA. It is not a bug: the constant-deflection criterion (15% drop) requires pressure to scale linearly with load — a heavier rider deflects the tire more at the same pressure. Calculators that optimize impedance saturate with weight; those that optimize deflection do not. We chose deflection because it is the measurable, reproducible criterion. And we are inside Berto's empirical domain: he measured up to 220 lbs per wheel. In practice, the hookless cap (72.5 PSI) compresses this divergence on most modern wheels.
Berto explicitly states the 15% criterion is not valid for MTB: proportionally narrower rims, knobs that distort the measurement, and an off-road objective (traction, absorption) different from rolling. Rather than invent numbers, we say so: the MTB module will ship separately, based on manufacturer heuristics (Maxxis, Schwalbe, Continental). The selector is already in the interface, disabled.
Berto measured with 1990s technology on 700C wheels. The casing correction is a 3-level approximation of a continuum. The two market references diverge from each other by up to 30% on gravel — there is no true "optimal pressure" to calibrate against, only reasonable empirical philosophies. This calculator picks one (constant deflection), declares it, and publishes its uncertainty. Adjust ±2-3 PSI by feel: the model is the rigorous starting point, not dogma.
Because they optimize different objectives. We apply constant deflection (Berto 15%), which scales linearly with load; SILCA optimizes impedance and saturates with weight; Rene Herse gives one value for both wheels. In cross-validation, 9 of 12 wheels landed within ±10% of the average of both — and every deviation has a documented physical cause in the methodology.
A 90% confidence interval via Latin Hypercube Sampling over your inputs' real uncertainty (weight ±2 kg, width ±1.5 mm, distribution ±3 points). It does not include your gauge error — that's your instrument's uncertainty. If you measure width with a caliper, the band shrinks to a third.
ETRTO limit, no exceptions (60 PSI if internal rim width is ≥30 mm). Exceeding it risks sudden blow-off. If your calculated value exceeds it, we show both and apply the cap. Never exceed the tire or rim manufacturer maximum — whichever is lower.
Berto himself states the 15% drop doesn't hold for MTB (proportionally narrower rims, knobs distorting the measurement, traction rather than rolling as the objective). The MTB module will ship separately with manufacturer heuristics — rather than invent numbers, we say so.
Ravello Joo, C. E. (2026). BikeLab PSI Calculator (Version 1.0.0) [Software]. BikeLab Studio. https://doi.org/10.5281/zenodo.20673124
Software deposited with a permanent DOI, indexed in OpenAIRE. Code can be copied; dated authorship cannot.